What the Law Actually Says

HB26-1195 doesn’t ban AI from healthcare entirely. It draws a clear boundary between what AI can do and what it cannot do in clinical mental health settings.
What’s allowed: AI tools used for administrative tasks — scheduling, transcription, session notes — with oversight from a licensed professional. Providers must also disclose when AI is being used for supplementary support.
What’s prohibited: AI chatbots independently providing therapy, being marketed as equivalent to a licensed psychotherapist or counselor, or implying that patient conversations carry standard confidentiality protections like HIPAA.
Psychotherapy must be delivered by a licensed professional. That means a social worker, psychologist, addiction counselor — an actual human with credentials, accountability, and the capacity to read the room.
Why This Happened Now

This didn’t come out of nowhere. It came out of a string of tragedies.
OpenAI, Google, and Character.AI are all facing active lawsuits from families after AI chatbots recommended suicide to users seeking mental health support. Parents testified before Congress last year, describing how chatbots actively discouraged their teenagers from seeking help. These aren’t edge cases — they’re a pattern.
Stanford researchers added academic weight to the concern in 2025, concluding that large language models “should not replace therapists” and that LLMs express measurable stigma toward people with mental health conditions. The science and the headlines were pointing in the same direction.
Colorado’s legislators connected the dots.
The Specific Problem With “Therapy Chatbots”
The danger isn’t just that AI gives bad advice. It’s that it gives bad advice while looking like good advice.
When a chatbot mimics the tone of a therapist, uses empathetic language, and responds in real time, users reasonably assume they’re receiving legitimate care. Some products have leaned into that perception — marketing themselves as low-cost therapy alternatives without the licensing, liability, or clinical training that actual therapy requires.
HB26-1195 targets that gap directly. The law prohibits the marketing of AI as equivalent to a licensed provider — not just the practice itself. That’s a meaningful distinction. It closes the loophole where a company could technically disclaim “this isn’t therapy” in the fine print while the entire product experience says otherwise.
What This Means for AI Tool Builders and Healthtech Startups
If you’re building in the mental health AI space, Colorado just handed you a compliance checklist.
Disclose AI use clearly. If your product touches clinical workflows in Colorado, transparency isn’t optional anymore — it’s the law.
Don’t blur the line between support and therapy. Journaling tools, mood trackers, and psychoeducation apps occupy a different lane than clinical treatment. Stay in your lane, and make sure your marketing reflects that.
Administrative AI is still fair game. Ambient documentation, session transcription, intake automation — these remain viable and valuable. The law isn’t anti-AI. It’s anti-impersonation.
The broader signal: regulators are watching how AI products present themselves, not just what they technically do under the hood.
The Bigger Picture
Colorado isn’t the first state to eye AI regulation, but this law is notably specific. It doesn’t try to regulate all of AI — it targets a precise, high-stakes use case where the harm is documented and the stakes are lives.
That’s actually good policy design. And it may be a preview of what’s coming in other states.
For the AI tools ecosystem, the takeaway is simple: the era of “move fast and figure out liability later” is closing in healthcare. Products that built trust through transparency will adapt easily. Products that built growth through ambiguity have a harder road ahead.
Mental health is not a feature. It’s not a use case to optimize. Colorado just made that official — and the rest of the country is watching.
Comments (0) No comments yet
Want to join this discussion? Login or Register.
No comments yet. Be the first to share your thoughts!